Add core components for request coalescing and service management

- Introduced coalescing logic in `coalescing.go` to handle concurrent identical upstream fetches, including a state machine and response buffering for improved performance.
- Implemented a new cache file format in `format.go`, supporting serialization and deserialization of HTTP responses, along with range request handling.
- Developed an HTTP handler in `handler.go` to manage requests, including special endpoint handling and metrics reporting.
- Added rate limiting functionality in `ratelimit.go` to control per-client and global request rates, enhancing security and performance.
- Created service management capabilities in `service.go` to define and manage cacheable services, including user-agent detection.
- Updated tests in `steamcache_test.go` to cover new functionalities, ensuring robustness and reliability across the codebase.

steamcache/coalescing.go

@@ -0,0 +1,117 @@
+// steamcache/coalescing.go
+// Request coalescing (de-duplicating concurrent identical upstream fetches for the same
+// cache key). Includes the coalescedRequest state machine + waiter/leader coordination,
+// response buffering for thundering herd avoidance, and the coalescer wrapper that
+// owns the in-flight map + mutex (SteamCache methods delegate; no direct map access
+// in core or handler).
+package steamcache
+
+import (
+	"net/http"
+	"sync"
+	"sync/atomic"
+)
+
+type coalescedRequest struct {
+	waitingCount atomic.Int32
+	done         bool
+	mu           sync.Mutex
+	// Buffered response data for coalesced clients
+	responseData    []byte
+	responseHeaders http.Header
+	statusCode      int
+	// Broadcast signal for all waiters (closed by leader in complete)
+	doneCh        chan struct{}
+	completionErr error
+	// Active protocol (post-legacy cleanup): waiters wake on doneCh, then read completionErr/response* under mu (or pre-unlock copies in waiter).
+}
+
+func newCoalescedRequest() *coalescedRequest {
+	cr := &coalescedRequest{
+		done:            false,
+		responseHeaders: make(http.Header),
+		doneCh:          make(chan struct{}),
+	}
+	cr.waitingCount.Store(1)
+	return cr
+}
+
+func (cr *coalescedRequest) addWaiter() {
+	cr.waitingCount.Add(1)
+}
+
+func (cr *coalescedRequest) complete(resp *http.Response, err error) {
+	cr.mu.Lock()
+	defer cr.mu.Unlock()
+	if cr.done {
+		return
+	}
+	cr.done = true
+
+	if err != nil {
+		cr.completionErr = err
+	} else {
+		// Store response data for coalesced clients
+		if resp != nil {
+			cr.statusCode = resp.StatusCode
+			// Copy headers (excluding hop-by-hop headers via filter)
+			cr.responseHeaders = filterHopByHopHeaders(resp.Header)
+		}
+	}
+	// Broadcast to *all* waiters (thundering herd fix). Close is safe here because of the done guard above.
+	close(cr.doneCh)
+}
+
+// setResponseData stores the buffered response data for coalesced clients
+func (cr *coalescedRequest) setResponseData(data []byte) {
+	cr.mu.Lock()
+	defer cr.mu.Unlock()
+	cr.responseData = make([]byte, len(data))
+	copy(cr.responseData, data)
+}
+
+// coalescer owns the coalesced requests map and mutex. It encapsulates the
+// in-flight request dedup state so SteamCache no longer directly manipulates
+// the raw map (Phase 2 extraction). Unexported; same-package access for tests.
+type coalescer struct {
+	mu       sync.Mutex
+	requests map[string]*coalescedRequest
+}
+
+// newCoalescer constructs an empty coalescer (called from SteamCache.New).
+func newCoalescer() *coalescer {
+	return &coalescer{
+		requests: make(map[string]*coalescedRequest),
+	}
+}
+
+func (c *coalescer) getOrCreate(cacheKey string) (*coalescedRequest, bool) {
+	c.mu.Lock()
+	defer c.mu.Unlock()
+
+	if cr, exists := c.requests[cacheKey]; exists {
+		cr.addWaiter()
+		return cr, false
+	}
+
+	cr := newCoalescedRequest()
+	c.requests[cacheKey] = cr
+	return cr, true
+}
+
+func (c *coalescer) remove(cacheKey string) {
+	c.mu.Lock()
+	defer c.mu.Unlock()
+	delete(c.requests, cacheKey)
+}
+
+// getOrCreateCoalescedRequest delegates to the owned coalescer (preserves
+// existing call sites in handler.go and any white-box tests unchanged).
+func (sc *SteamCache) getOrCreateCoalescedRequest(cacheKey string) (*coalescedRequest, bool) {
+	return sc.coalescer.getOrCreate(cacheKey)
+}
+
+// removeCoalescedRequest delegates to the owned coalescer.
+func (sc *SteamCache) removeCoalescedRequest(cacheKey string) {
+	sc.coalescer.remove(cacheKey)
+}